For thirty years, Arm made money by letting other companies build chips using its designs. On Wednesday, it started building its own.
Arm Holdings unveiled the AGI CPU at its "Arm Everywhere" event in San Francisco, marking the company's first in-house processor designed specifically for AI workloads in data centers. Shares jumped more than 18% on the news, with Meta Platforms signing on as the lead customer and additional commitments from OpenAI and Cloudflare.
What the AGI CPU Actually Is
The chip is built on Taiwan Semiconductor Manufacturing Company's advanced 3-nanometre process and is optimized for energy efficiency. Arm says it delivers roughly double the performance-per-watt compared with conventional x86 processors from Intel and AMD - a meaningful claim in a data center market where power consumption is one of the central cost challenges for AI operators.
The launch is a significant departure from Arm's traditional business model. The company has generated revenue for decades through royalties on chip designs licensed to companies including Apple, Nvidia, Amazon, and Google. Moving into full chip production means Arm is now competing - at least partially - with some of its most important customers.
The Revenue Projections Are Ambitious
Arm's own forecast for the chip business is aggressive. The company projects approximately $1 billion in revenue in fiscal 2027-28, rising to $2 billion by 2029 and potentially $15 billion by 2031. Bank of America estimates the move could expand Arm's total addressable market from roughly $2.4 billion in royalties in fiscal 2026 to as much as $100 billion in CPU-related opportunities by fiscal 2031.
The long-term earnings potential Arm is projecting comes in around $9 per share by 2031. Bank of America is more conservative, modeling closer to $6.50 per share based on lower assumptions for market size, growth, and competitive share.
The Analyst Pushback
The 18% share jump reflects genuine excitement about Arm's strategic expansion. The analyst skepticism reflects the reality of what it is walking into.
Bank of America flagged several near-term risks. The CPU market is becoming increasingly crowded, with established semiconductor players and hyperscalers developing their own in-house chips simultaneously. The bank noted that Meta and OpenAI - Arm's own launch customers - already have existing CPU agreements with AMD and Nvidia, which limits how much share Arm can realistically capture even among its own partners.
The broader concern is structural. The bigger AI grows, the more pressure Arm's core smartphone and consumer markets face from limited memory supplies. Global smartphone unit volumes are expected to decline 10% to 20% in 2026, which puts additional strain on the licensing revenue base that has funded Arm's growth to date.
What This Means for the AI Chip Market
Arm's entry into full chip production adds another significant player to a semiconductor landscape already under pressure from GPU shortages, 18-to-24-month equipment lead times, and aggressive in-house development programs at every major hyperscaler. Nvidia remains dominant in AI training workloads, but the inference and CPU-side of AI infrastructure is increasingly contested.
For business leaders evaluating their AI infrastructure strategy, Arm's AGI CPU represents a genuine new option - particularly for operators prioritizing energy efficiency at scale. Whether the company can execute on its projections in a market this competitive is a different question entirely. The launch is real. The $15 billion revenue target by 2031 has a long way to go to prove itself.
